JPS62270712A - System for detecting condition of blast furnace - Google Patents

Abstract

PURPOSE:To forecast channeling or slip in a blast furnace by forming various date showing the conditions of the furnace with date on the furnace from a data inputting means, forming true and false data and performing inference and operation by artificial intelligence on the basis of the true and false data and information bases. CONSTITUTION:A time series processing means 16 for artificial intelligence, a time series filing means 17 for artificial intelligence, a sensor data preprocessing means 18 and an interface buffer 19 are incorporated into a conventional large-sized computer 10 so as to preprocess various sensor data 11 for operation with a small-sized computer 20. The computer 20 is composed essentially of information bases 21, 22, 23, a common data buffer (BB= blackboard) 24 and an inference engine 25. The information bases 21, 22, 23 are an information base 1 for diagnosing abnormal conditions of a furnace, an information base 2 for deciding the temp. of the furnace and an information base 3 for action to the temp. of the furnace, respectively. Results inferred by the inference engine 25 are displayed on CRT30.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a blast furnace condition detection system, and particularly to a blast furnace blow-through and slip estimation system.

[Conventional technology]

Conventionally, as a method for diagnosing and managing blast furnace conditions, blast furnace operators have qualitatively judged information from various sensors installed in the blast furnace, evaluated the blast furnace condition, and determined the optimal operating factors. However, the evaluation results vary depending on the ability and experience of the operator, making it difficult to standardize operational actions, and the evaluation is not quantitative, making operational analysis difficult. The problem was that it was difficult to do.

For this reason, a method for detecting blast furnace conditions has been proposed, for example, as disclosed in Japanese Patent Application Laid-Open No. 59-84705. This blast furnace status detection method selects factors that are known to be important from experience from among various sensor information,
By correlating and quantifying these with the phenomena inside the furnace, and by examining these organization and quantification from both short-term and long-term perspectives, it is possible to detect the status of the blast furnace. Appropriate situation management has been achieved.

[Problem that the invention seeks to solve]

The conventional method for detecting blast furnace conditions disclosed in Japanese Patent Application Laid-Open No. 59-64705 uses information from sensors.

The information is input into an analytical model and predetermined calculations are performed. For this reason, the computer that executes the calculation uses, for example, Fortran as the language, but the calculation capacity is extremely large. Furthermore, since blast furnaces change over time, it is necessary to change the analysis model itself for maintenance, but since the analysis model itself is complex, changing the conditions of the analysis model is an extremely troublesome task.

The present invention was made in order to solve these problems, and provides a blast furnace condition detection system that has a small calculation capacity when implemented on a computer, and that can be easily changed as the blast furnace ages. The purpose is to obtain.

[Means for solving problems]

The blast furnace condition detection system according to the present invention includes a data input means that takes in data from various sensors installed in the blast furnace at a predetermined timing, and a data input means that takes in data from various sensors installed in the blast furnace, unloading speed, pressure loss, shaft pressure, and shaft temperature based on the data from the sensors. ,
Means for creating various data indicating the status of the blast furnace such as fixed sonde temperature, gas utilization rate, furnace mouth sonde temperature, etc., means for creating truth/false data by comparing each of the above-mentioned verification data with its reference data, and truth/false data. , a knowledge base means storing knowledge bases of W based on experience and achievements regarding blast furnaces, and based on the truth data of the storage means and the knowledge base of the knowledge base means. The apparatus is equipped with an inference means for making a predetermined inference and predicting an atrium or slip.

The means for creating various data and the means for creating truth/false data described above are for obtaining truth/false data of the blast furnace necessary for performing inference calculations by the inference means, and serve as a preprocessing calculation function. There is.

[Effect]

In the present invention, after the blast furnace data from the data input means is used to create various data indicating the status of the blast furnace, truth data is created, and inference calculations are performed as an artificial intelligence based on the truth data and the knowledge base. , the stairwell or predict a slip.

〔Example〕

Embodiments of the present invention will be described below based on the drawings.

Figure 1 is a conceptual diagram of a blast furnace status detection system according to an embodiment of the present invention.
It is a large computer used for control, etc., and includes a time series processing unit 19 (L5, time series file unit αJ, and system processing unit Q4) that inputs and processes data from various sensors αυ in time series. , these have the same configuration as conventional detection systems.In this embodiment, each of the above devices is equipped with a time series processing means (for artificial intelligence).
Ql, time series file means (for artificial intelligence) αη, sensor data preprocessing means OB, and interface buffer Q91 are incorporated. These devices surrounded by dashed lines in the figure perform preprocessing of sensor data for calculation by a small computer, which will be described next.

(branch) is a small computer, knowledge base 1 (for abnormal reactor condition diagnosis 1c!ll), knowledge base 2 (for furnace heat judgment) (company), knowledge base 3 (for reactor heat action) (23, common data buffer (to) and an inference engine (sha).

(To) is a CRT that displays the results of inference by the inference engine.

It should be noted that, in the large-scale computer α0) of FIG. 2, the parts corresponding to the constituent parts shown by broken lines in the conceptual diagram of FIG. 1 are mainly shown. Sensor (l1m), (l
lb) and (lie) correspond to the various sensors αυ in FIG. 1, and the sensors include all sensors installed in conventional blast furnaces, such as temperature sensors, pressure sensors, and gas sensors for blast furnaces. (41) is an interface, (42)
is the CPU, and (43) is the ROM in which the program is stored.
, (441, (45) are RAM, (46) is interfunis. CPU (42) and ROM (
43) is based on the program stored in the
The time series processing means side and sensor data preprocessing means (
). The RAM (44) constitutes the time series file means α in FIG. RAM (4
5) is a storage means for temporarily storing sensor data that has been subjected to preprocessing, which will be described later, and constitutes the interface buffer shown in FIG. 1 together with the interface (46).

In the small computer (■) in Figure 2, (47) is the keyboard, (48) is the interface, (49) is the CPU, (50) is the ROM, and (51) to (54) are the RA.
M, (55) is an interface. CPU (
49) and ROM (50) constitute the inference engine means (company) of FIG. 1 based on the programs stored therein. The RAM (51) constitutes the knowledge base 11211 in FIG. 1, and the RAM (52) constitutes the 9 [Pace (Inc.). These RAMs (51), (52
) may be constructed from a ROM if its storage contents are fixed.

The RAM (53) constitutes a knowledge base (a), and when the system builder wants to change or add to its memory contents, he inputs it using the keyboard (47) and then inputs it via the interface (48). Memorize the contents. R.A.
M (54) is the common data buffer means (to) in FIG.
It consists of a large computer RAM (4
5) is stored via the interface (46). The results calculated by the CPU (49) are sent to the CRT G[
) is displayed. Note that this embodiment shows an example in which the time-series processing means side, the time-series file means a7+, and the sensor data pre-processing means (to) are incorporated into a large-scale computer 0ω, but this is similar to the existing large-scale computer GO). The aim is to make effective use of the remaining capacity, so it goes without saying that these can be incorporated into a small computer.

The operation of this embodiment having the above configuration will be explained with reference to the flowchart in FIG. 3 and the explanatory diagram in FIG. 4.

(11) First, data from various sensors αD are sequentially read at predetermined timing by the time series processing means and stored in the time series file means (step 31).

Specifically, this operation is performed by the sensors (Ilal, (Ilb),
(llc)- data is stored in the RAM (44) via the interface (41).

(2) The data stored in the time series file means αη is processed by the sensor data preprocessing means αg (step 32). This operation is performed by the CPU (42) in FIG. The contents of sensor data preprocessing will be specifically explained below.

This sensor deck preprocessing includes data processing regarding unloading, pressure loss, temperature, gas utilization rate, and tap slag.

(a) Unloading; (a-1,) Unloading speed Vi (i-1 to 4) ■ Data within the specified time (:=30 seconds) from landing on the floor is not processed.

■ From then on, every minute ■ Calculation during winding If the time between the previous calculation and winding is less than 1 minute, calculate Vi as follows.

■If Vi is between ■ and ■ above, use the latest Vi calculation value. However, when it is outside the range of Vmin-◇maχ, the previous Vi calculation value is used. (Excluding data such as slippage) (a-2) Variation in unloading speed σVI (i-1~
4) (time, t"-n, and the calculation is based on the latest hour (n = 59)
.

(a-3) Amount of unloading delay Vv1 Directly, however, (a) Vo will not be released until the current time (k-0) is reversed.
(If not, find the point 1 hour ago).

(b) The point where it became below vlE (k-')
(If the value is less than vlE from 1 hour ago, all values are integrated.)

vo: Theoretical descent speed ΔVβ: Set value Vs = V6 - ΔVβ Theoretical descent speed: vo CB O/C, CB □+□ CB: Coke base (coke amount per 16 hours) ρ
ζ: Coke bulk density 07C: Ore/coke ratio.

ρ. 2M stone average bulk density Vβ: Blow flow rate V, /, Blow unit...Setting (threshold) Pig amount: Planned iron extraction amount/eh S: Hearth cross-sectional area k: Correction coefficient (a-4 ) Average unloading speed Vi Direct (b) Pressure loss; (b-t) Pressure loss (air permeability) k, , '(a) xb (Note) Hot-blast stove switchability is not calculated.

Pl: Blowing pressure P,: Furnace top pressure vbo Soh: Bosch gas velocity (B) 0' to Kb- >ΔKb+ (#=1.2)
When determining '=, 0. Normal time = K, -f (ΔVχ) 2 When the wind is reduced (ΔVχ)・Air reduction amount (b-2) Daily average pressure drop (7'01' to 7'00'
) K.

However, (a) Kbt during hot stove switching (=pressurizing) and air reduction are excluded from the calculation.

(if) The calculation result is within the specified range (Kmin~Kmax
) When going outside, use the previous day's ni・.

(b-3) Variation in pressure drop (latest 1 hour) σkb
(Note) Excludes from calculation while hot air stove is being switched.

(b-4) Average shaft pressure P1J in steady state (
i=1-4. j-1 to 10:40); 1 hour average value (
(X'01' to X+1°00') However, whether it is stable or unstable is determined from the shaft pressure fluctuation factor of the pilot system, and if it is stable, the current one-hour average value is used, and if it is unstable, the same value as the previous time is used.

(C) Temperature; (C-1) Shaft temperature increase (C-2) Steady state shaft temperature average Ti j
(i-1 to 8, 1 to 4. j-1 to 740): 1 hour average value (X'01' to X+1'00') However, X'0
1' to X+1'OO' +7), the previous TI is used as is.

(C-3) Fixed sonde temperature increase (C-4) Fixed sonde temperature average T'1j (i
"1~8.1~16.j-1~2:2); 1 hour average value (X@01'~X+1000') However, X@0
Even if there is even one change in dT'ij □□□≧100℃73 minutes between 1' and X+1"oo' (Please use the previous TNj as is.

(d) Gas utilization rate; (d-1) r) CO descent 4?゛°-9...(Current value)-7...(3 minutes ago) t rico=CO2/ (co+co,)(d-2)
1 co falling amount Δηe.

Δ+7eo"+7co (current value) - ηco (value 9 minutes ago) (e) Tapping slag; (e-1) Hearth sonde temperature center (1) (e-2)
Center of furnace mouth sonde temperature (2) Tc' (current value) - Tc'
(value of minutes ago) ∆TeTc': Including current value, past 3
Current estimated value by linear regression of Tc data for 0 minutes (e-3) Around furnace sonde temperature (11 (e-4)
Around the furnace mouth sonde temperature (2) Te' (current value) -Te
' (value of the previous minute)>ΔTeTe': Find it using the same calculation as Te'.

(e-5) Daily average air blowing pressure (a) Pb (average of 7"01' to 7.oo') Exclude from calculation during 4 conversion, hot blast stove switching, and air reduction. Calculation result is P1 When outside the range of in~P4m+mouth, use the previous P as is.

(a) When Wb Pb'>ΔP-;
b: Normal snow P, -f' (ΔVX): Determine h) - (e) in the same manner as above when the wind is reduced. The indicated predetermined truth data is created and stored in the interface buffer αl. Specifically, it is stored in RAM (45) in FIG.

(3) Next, the truth data stored in the interface buffer is transferred to the common data buffer (2) (S3). Specifically, the data stored in the RAM (45) in FIG. 2 is transferred to the RAM via the interface (46).
(54).

(4) The inference engine means (company) uses the knowledge data stored in advance in the knowledge base 12υ and the common data buffer (
The situation inside the blast furnace is inferred based on the truth/false data of 2) (S4). Specifically, the data of RAM (51) and R
Based on the data of A M (54), CPU (4
9) performs a predetermined calculation.

Here, the knowledge base 1ell is composed of knowledge units as shown in FIG. 4, taking into account the efficiency of inference, and this figure shows the knowledge base for the atrium. These are mainly based on the operator's knowledge and experience.
It is expressed as an EN~ type production rule, and each rule knowledge unit is directly connected to the atrium and
There are two units: one that determines slippage and one that determines the state of the iron slag, which has a major impact on blow-through and slippage.
It consists of two groups, and as a whole, it determines the degree to which blow-through/slip abnormalities occur. In FIG. 4, OR is not a normal logical sum, but a sum in which the CF-value is taken into account for each truth/false data, and AND is a normal logical product. For example, in the unloading rules, the CF By calculating the sum of these values in consideration of the values, data indicating that "unloading is out of order" along with the CF value is obtained. Similarly, in the pressure-related rules, based on the true/false data of ``the pressure loss is large'', ``the pressure loss has a large variation'', and ``the shaft pressure is high'', these rules are set in consideration of the CF value. The sum is calculated, and data indicating that "pressure fluctuation is large" is obtained along with the CF value. For temperature-related rules, the following truths are true: "The shaft temperature is rising rapidly,""The amount of increase in shaft temperature is large,""The fixed sonde temperature is rising rapidly," and "The amount of increase in fixed sonde temperature is large." Based on the data, data indicating ``if the temperature fluctuation is large'' along with the CF value can be obtained in the same manner as above. In the gas utilization rate related rules, "the gas utilization rate is rapidly declining"
Based on the true/false data that "the amount of decrease in the gas utilization rate is large", data indicating that the gas utilization rate is decreasing is obtained along with the CF value. “There is a lot of residue (sensor judgment)
In inference based on sensor information as to whether
The logical product of the true/false data of "The center temperature of the furnace mouth sonde is decreasing" and "The surrounding temperature of the furnace mouth sonde is rising" is taken, and if either one is "1", then the The logical product is "false".

Except for this point, the case is the same as the above case. In addition, in other residue (sensor) related rules, true/false data such as "shaft pressure has increased by n or more", "air blast pressure has increased rapidly", and "residue pig iron amount is large" are used. Furthermore, the rules include a [human judgment rule] that incorporates the operator's judgment, and data can be obtained as to whether there is "a lot of residue (human judgment)" or not.

Next, the above data, that is, ``unloading is irregular'',
The sum of the data for "pressure fluctuation is large", "temperature fluctuation is large", and "gas utilization rate is decreasing" is calculated by taking into account the CF value, and the "atrium" ( (sensor judgment)” data can be obtained. Similarly, "There is a lot of residue (sensor judgment)" and "There is a lot of residue (sensor judgment)"
The sum of these data is calculated by considering the CF value for the data regarding ``human judgment)'', and the sum of these data is calculated by considering the CF value.
” can be obtained.

Then, the sum of the above-mentioned "Atrium (sensor judgment)" and "Atrium (residual judgment)" as well as "Atrium (previous warning)" is calculated by considering the CF value, and "Atrium (overall judgment)" is made.

(5) Next, the results of the above comprehensive judgment are displayed on the CRT OQ. That is, the calculation result in the CPU (491) is output to the CRT (71) via the interface (55) and displayed (step S5).

(6) The presence or absence of a stop signal is determined. If "present", the calculation is stopped; if "absent", the process returns to step 1 (Sl) (step S6).

The above steps (Sl) to (S6) are repeated at predetermined time intervals (for example, every two minutes).

The above explanation is for the case of an atrium, but it is basically the same in the case of a slip, and the CF in the case of a slip is
The only difference is that the value is different from the CF value in the case of open ceiling. Note that the part surrounded by the broken line in Figure 4 is common to both atriums and slips.

Atrium CF value and slip CF obtained as above
A time chart of the values and unloading speed is shown in FIG. This diagnostic result has shown an extremely good track record, achieving an accuracy rate of over 85% for slips.

Note that knowledge base 1X5 and knowledge base 3 (2
) are not directly related to the present invention, so here we will only point out that the basic idea of the present invention can be applied to furnace heat determination and furnace heat action, and will omit detailed explanation thereof.

〔Effect of the invention〕

As described above, the present invention creates true/false data from various data installed in a blast furnace, and uses this data and a knowledge base based on experience recorded in a knowledge base means as an artificial intelligence system. Since the predetermined inference is made, conventional experience can be fully utilized, and even if the system is implemented on a computer, its capacity will be extremely small.

Furthermore, changes in the experience of the blast furnace can be made only by changing the stored contents of the knowledge base means, making it extremely easy to make changes.

[Brief explanation of the drawing]

Fig. 1 is a conceptual diagram showing a system according to an embodiment of the present invention, Fig. 2 is a block diagram showing the eight-domain configuration of the conceptual diagram in Fig. 1, and Fig. 3 shows the operation of the system in Fig. 1. FIG. 4 is an explanatory diagram showing the configuration of the knowledge base and its inference process, and FIG. 5 is a time chart showing the output (judgment result) of the system shown in FIG. Agent Patent Attorney Tadashi Sato Figure 2

Claims (1)

[Scope of Claims] Data input means that takes in data at predetermined timing from various sensors installed in the blast furnace, unloading speed, pressure loss, shaft pressure, etc. based on the data from the sensors;
Means for creating various data indicating the status of the blast furnace such as shaft temperature, fixed sonde temperature, gas utilization rate, furnace mouth sonde temperature, etc., means for creating truth/false data by comparing the various data with reference data, true A storage means for temporarily storing false data, a knowledge base means for storing various knowledge bases based on experience and achievements regarding blast furnaces, and a knowledge base based on the truth/false data of the storage means and the knowledge base of the knowledge base means. A blast furnace situation detection system characterized by comprising an inference means for making a predetermined inference and predicting blow-through or slip.